In February 2026, Dakan and Feller published an AI Fluency Index built from 9,830 Claude conversations, classifying 11 observable collaboration behaviors across three axes: Description, Discernment, and Delegation. I ran their taxonomy against my own Claude Code sessions and found I was hitting maybe four of the eleven with any regularity. The rest were behaviors I'd read about but never actually triggered in practice.
That gap is what skill-tree is for.
The tool installs as a Claude Code plugin — claude plugin marketplace add robertnowell/ai-fluency-skill-cards && claude plugin install skill-tree-ai@ai-fluency-skill-cards — and on each session analyzes your message history against those same 11 behaviors. It assigns one of seven archetype cards (rendered as tarot cards with museum art, live example at skill-tree-ai.fly.dev/fixture/illuminator) and designates one behavior you haven't touched as a growth quest for your next session. That quest persists via a SessionStart hook so Claude surfaces it when you open a new conversation.
The classification runs remotely on Fly.io in 30–60 seconds: extract user messages, classify against the fluency framework, assign archetype, synthesize narrative, render, return a stable URL.
The archetype isn't the point. The point is the gap list — behaviors you're systematically skipping without knowing it. Seeing "Discernment" axis behaviors at near-zero across 40 sessions is harder to ignore than a vague sense that you might be in a rut.
Also available as an MCP server (npm install skill-tree-ai) for Cursor, VS Code, and Windsurf.
github.com/robertnowell/skill-tree
Top comments (0)